Proof of Theorem 4.6, even case — make-span against
ProvedAlgMechDesign.LowerBound.ratio_step_evenmechanism-designp2o-batch-p200bp2o-gran-per-chapterp2o-plan-paperp2o-v1scheduling
In the setting of Claim 4.8 (two agents, tasks, a truthful direct mechanism , with , , and ), assume that is even. Then the mechanism's make-span at equals , while some allocation does much better:
(The allocation meant is the one giving agent 1, in addition to , half of agent 2's original tasks.) Letting this gives the ratio in the lower bound.
Formalization Note Agents 1 and 2 are 0 and 1 in Fin 2. The good allocation is existentially quantified. The odd case of the paper's proof is not part of this statement.
Preamble
import Mathlib import Definitions.Def_AlgMechDesign_LowerBound_Model
Formal statement
namespace AlgMechDesign.LowerBound
/-- The ratio step of the proof of Theorem 4.6 (even case): in the setting of Claim 4.8, if
`|x²(t)|` is even then at `t̂` the mechanism's make-span is `|x²(t)|`, while some allocation has
make-span at most `|x²(t)|/2 + k·ε`. -/
theorem ratio_step_even {k : ℕ} (hk : 3 ≤ k) (alloc : (Fin 2 → Fin k → ℝ) → (Fin k → Fin 2))
(pay : (Fin 2 → Fin k → ℝ) → Fin 2 → ℝ) (htruth : IsTruthful alloc pay)
(t : Fin 2 → Fin k → ℝ) (ht : t = fun _ _ => 1)
(hcard : (taskSet (alloc t) 0).card ≤ (taskSet (alloc t) 1).card)
(heven : Even (taskSet (alloc t) 1).card)
(x : Finset (Fin k)) (hx : x = taskSet (alloc t) 0)
(ε : ℝ) (hε0 : 0 < ε) (hε1 : ε < 1)
(that : Fin 2 → Fin k → ℝ)
(hthat : that = Function.update t 0 (fun j => if j ∈ x then ε else 1 + ε)) :
makespan that (alloc that) = ((taskSet (alloc t) 1).card : ℝ) ∧
∃ y : Fin k → Fin 2,
makespan that y ≤ ((taskSet (alloc t) 1).card : ℝ) / 2 + (k : ℝ) * ε := by sorry
end AlgMechDesign.LowerBound
Source
Nisan, Ronen, Algorithmic Mechanism Design, Games Econ. Behav. 35, 2001, p. 179, proof of Theorem 4.6, paragraph after Claim 4.8
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.